Oecologia (2001) 126:216–224 DOI 10.1007/s004420000516 Martin Predavec · Charles J. Krebs · Kjell Danell Rob Hyndman Cycles and synchrony in the Collared Lemming (Dicrostonyx groenlandicus) in Arctic North America Received: 11 January 2000 / Accepted: 21 August 2000 / Published online: 19 October 2000 © Springer-Verlag 2000 Abstract Lemming populations are generally character- Introduction ised by their cyclic nature, yet empirical data to support this are lacking for most species, largely because of the Lemmings are generally known for their multiannual time and expense necessary to collect long-term popula- density fluctuations known as cycles. Occurring in a tion data. In this study we use the relative frequency of number of different species, these cycles are thought to yearly willow scarring by lemmings as an index of lem- have a fairly regular periodicity between 3 and 5 years, ming abundance, allowing us to plot population changes although the amplitude of the fluctuations can vary dra- over a 34-year period. Scars were collected from 18 sites matically. The collared lemming, Dicrostonyx groen- in Arctic North America separated by 2–1,647 km to in- landicus, is no exception, with earlier studies suggesting vestigate local synchrony among separate populations. that this species shows a strong cyclic nature in its popu- Over the period studied, populations at all 18 sites lation fluctuations (e.g. Chitty 1950; Shelford 1943). showed large fluctuations but there was no regular peri- However, later studies have shown separate populations odicity to the patterns of population change. Over all to be cyclic (Mallory et al. 1981; Pitelka and Batzli possible combinations of pairs of sites, only sites that 1993) or with little or no population fluctuations (Krebs were geographically connected and close (<6 km) et al. 1995; Reid et al. 1997). Differences in fluctuations showed significant synchrony in fluctuations. The popu- found in different populations lead directly to the ques- lations studied may not even be cyclic, at least for the tion of geographic synchrony. Krebs and Myers (1974) time period 1960 to 1994, and although fluctuating, ran- suggest that lemming populations are synchronous domisation tests could not reject the null hypothesis of across large geographic areas and this has been support- random fluctuations. These data have implications for ed largely by data from northern Europe (e.g. Henttonen the testing of hypotheses regarding lemming cycles and et al. 1977; Myrberget 1973). A number of studies highlight the need for long-term trapping data to char- have, however, shown a general lack of synchrony (e.g. acterise the lemming cycle. Myrberget 1973), including studies from North America (Pitelka and Batzli 1993). Keywords Lemming cycle · Synchrony · Dicrostonyx The question of synchrony has clouded much of the groenlandicus debate regarding the relative merits of various hypothe- ses proposed to explain lemming cycles. Many intrinsic M. Predavec (✉) · C.J. Krebs hypotheses (e.g. the Chitty hypothesis, Chitty 1960) at Department of Zoology, University of British Columbia, their core would suggest asynchrony among geographi- 6270 University Blvd., Vancouver, B.C. V6T 1Z4, Canada cally separated populations. However, synchronising fac- e-mail: [email protected] Tel.: +61-3-99055652, Fax: +61-3-99055613 tors such as weather (Chitty 1996) are often included in order to explain apparent geographic synchrony, creating K. Danell difficulties in formulating precise, testable predictions. Department of Animal Ecology, Swedish University of Agricultural Sciences, 901 83 Umeå, Nearly all past work on lemming cycles suffers from Sweden a common problem of ecology, namely a lack of long term data. Often statements are made regarding popula- R. Hyndman Department of Mathematics, Monash University, Clayton, VIC, tion fluctuations based on time series of less than Australia, 3168 10 years. For populations that are thought to have a cy- Present address: clic periodicity of 3–5 years, such data allow for only M. Predavec, Department of Biological Sciences, two, possibly three, complete cycles, and time series of Monash University, Clayton, VIC, Australia, 3168 at least 30 years are necessary for rigorous analysis. 217 Studies containing long time-series tend to use indirect estimates of lemming abundance. For example, Elton (1942) used the sales of Arctic fox skins over a 53 year period as an estimate of lemming abundance and Steen et al. (1990) analysed a 79-year time series based on re- ports of hunters, foresters and locals of the abundance of lemmings. However, in the case of using fox skins this covers very broad geographic areas and may be affected by variations in the fox populations independent of lem- ming fluctuations. Similarly, using anecdotal evidence is open to potential bias. In this paper we use an indirect method to record for- mer fluctuations in populations of the collared lemming (Dicrostonyx groenlandicus). During winter months, while living in subnivean spaces, collared lemmings eat the bark of willow stems (Batzli 1993). As long as stems are not ring-barked they will continue to grow and lay down yearly growth rings, apart from at the point where bark was removed, hence creating a scar. By taking a cross-section through a living stem at the point of the scar and counting growth-rings, it is possible to date the winter during which the bark was removed (Danell et al. 1981). Based on the assumption that when there are more lemmings there will be more scars, changes in the relative frequency of scars among years will reflect changes in the lemming population. The validity of this assumption has been supported by studies looking at temporal variation in vole numbers (Danell et al. 1981) and others looking at spatial variation in Siberian lem- ming populations (Danell et al. 1999; Erlinge et al. 1999). To test this assumption for populations of D. groenlandicus, in this study we compare the phases of the population fluctuations as derived from the frequen- cy of scars with available trapping data. We then apply the dendrochronological analysis to 17 sites in the Cana- dian high Arctic and one site in Alaska. The resulting da- ta allow us to test for periodicity and geographic syn- chrony in the lemming cycle over relatively long time Fig. 1 A Map showing Northern Canada and Alaska. A is the lo- periods and large geographic areas. cation of Deadhorse. The rectangle shows the magnified area in map B. B Map showing the main study locations in the Canadian High Arctic. B Byron Bay; C Cambridge Bay; D Turnagain Point; E Walker Bay Camp; F Old Den; G Walker Bay South; H Cock- Materials and methods burn Island; I Hurd Island; J Hope Bay; K Jameson Island; L Wil- mot Island; M Breakwater Island; N Fishers Island; O Banks Pen- Patterns of past population change were estimated at 18 sites insula; P Algak Island; R Ekalulia Island; S Bay Chimo (Fig. 1). Sites included mainland, peninsula and island locations and were chosen primarily for their ease of access, but some were chosen for specific comparisons, such as between island and near- Scars were collected during the summers of 1994 and 1995, ei- by mainland sites (e.g. Fig. 1, sites O and P). At each of the 18 ther at the start of summer, before the yearly growth of willows sites, estimates of past lemming population change were made us- had started, or towards the end of summer after the yearly growth ing the techniques of Danell et al. (1981). Approximately 300 period, allowing us to distinguish precisely in which winter scars scarred willow stems were collected from each site in areas of dry were made. hummocky tundra with a high cover of willows, the preferred hab- In the laboratory, scars were examined under a dissecting mi- itat of D. groenlandicus (Morris et al. 2000). A sample of 300 croscope and assigned to one of two categories based on their ap- scars gives a good representation of the past population patterns pearance. Many scars had rodent teeth marks still visible and these (K Danell, unpublished data). Scars were found by haphazardly were assigned to the “sure” category. Scars that were not clearly moving through a patch of Arctic willows (Salix lanata), with the created by rodents were assigned to the “unsure” category. Stems scars easily seen on stems. Stems were cut so that the entire scar were discarded if a rodent clearly did not create the scar, for ex- was included in the section of stem for closer examination in the ample if a branch had been pulled off. Seventy percent of stems laboratory. Stems were stored in a mesh bag and air dried before were assigned to the “sure” category. Using an electric band saw, being taken to the laboratory. Salix lanata usually has numerous stems were cut across the grain, through the scar. If more than one stems arising from the soil and so the sampling did not kill the scar was present on a single stem then these were kept together plants. during the following process. Cut stems were placed in warm wa- 218 ter for about 5 min to soften the wood and then the cut surface of cance level determined by a randomisation test based on 1,000 the stem was cleaned using a scalpel blade. Stems were allowed to random permutations of the original series. dry for at least 12 h before the prepared cut surface was smeared with a thin film of yellow-coloured zinc cream, which highlighted the growth rings. Using a dissecting microscope we counted the Testing for cyclicity number of growth rings laid down since the scar was made. Aged stems were assigned to either a “sure” category if the number of Based on the binary time series, if the data showed cyclicity rings was clearly visible, or an “unsure” category if the age of the we expected a regular sequence of 1 s and 0 s.